Walmart sales prediction based on random forest model and application of feature importance

Abstract

Sales forecasting is crucial for efficient resource allocation and inventory management in retail. This study employs Random Forest to predict weekly sales for 45 Walmart stores, leveraging a diverse dataset with store-specific sales and external factors. Through meticulous preprocessing and model application, one achieves outstanding accuracy, with a Weighted Mean Absolute Error (WMAE) as low as 1.2030 and an impressive accuracy rate of 98.8%. Additionally, integrating feature importance ranking sheds light on influential variables in sales forecasting. This study provides a blueprint for developing precise and adaptable sales forecasting models, offering profound significance for the retail industry. It underscores the effectiveness of machine learning techniques, e.g., Random Forest and insightful feature engineering in achieving highly accurate predictions. By enhancing the industry's understanding of intricate sales dynamics, this research contributes to optimizing resource allocation, inventory management, and strategic planning. Ultimately, it drives operational efficiency and success in the dynamic landscape of the retail sector.